Introduction

Because I’m Happy…? (Valence density plot)

• You can clearly see that most of my favorite songs in 2016, 2017 and 2018 were “happier” than my most favorite songs in 2019 and 2020. Although 2019 has some outliers with music used for sleeping which all have a really low valence score.

Multivariate Musicology (Valence, energy, danceability)

• My music from 2016-2020 has generally become more and more danceable every year shifting to the right until in 2020 almost all songs have a danceability higher than 0.4. The energy distribution stays more or less the same every year.

• As mentioned in the previous tab you can see the music used for sleeping in the 2019 plot at the bottom left it has a clear low score on just about every variable.

Time (Tempo histogram with density)

• The tempo of the songs in my spotify wrapped playlists mostly seem to hang around 100-120 BPM this seems to conform to Moelants(2002) which suggests that humans prefer tempi around 120 BPM. Starting at 110 BPM in 2016 then 100 BPM in 2017/2018, shooting up to 120 BPM in 2019 and reverting back to 100 BPM in 2020.

And Time again (Tempograms for Favorites/Outliers)

Acchordingly (Chordograms for Favorites/Outliers)

• Outliers:

• 2016 Jesus of Suburbia - Green Day: an outlier because the song has lots of different sections and tempo changes.

• 2017 Weightless part 1 - Marconi Union: an outlier because its sleep music and is not something you would listen to on the regular.

• 2018 Bohemian Rhapsody - Queen: an outlier because it has different sections and tempo changes.

• 2019/2020 Build God, Then We’ll Talk - Panic! At The Disco: an outlier because it has different sections and tempo changes.

Keyp on moving (Key Frequency)

• The key frequency doesn’t change that much through the years other then flattening out a bit in 2017, 2019 and 2020. The mode seems fairly distributed between major and minor in most years except for 2018 where you can clearly see i listened to more songs with a major mode.

Feels like summer (Chromagram)

• Chromagram for the song i most frequently listened to. (under construction)

Self-Similarity Matrices

• Chroma and Timbre self-similarity matrices for the song i most frequently listened to. At multiple segments in the song (at around 120, 300 and constantly during the outro) there is a section where most instruments fade out and come back in with Childish Gambino belting a high note, this could explain the intersections at those times in the chroma an timbre graphs. Whats interesting is that the timbre graph singles out the segment at 200 seconds. This is because not just some but all the other instruments stop for a moment as a opposed to earlier and later occurrences of a similar segment, resulting in the biggest difference in timbre.

Wrapped Classification (KNN, Random Forests)

• 2019 was excluded due to errors.

• Overall the classifiers don’t work very well for my corpus but i expected as much. Since i do listen to a lot of different music the algorithms might have trouble finding distinct features for every year the different years. However, with limited features on the random forest classifier 2016, 2017 and 2020 seem to be working a lot better, almost classifying 50% correctly. Another thing worth noting is the correlation between 2016 and 2018 for both classifiers. The classifiers keep labeling 2016 as 2018 and vice versa. The k-nearest neighbors classifier was trained on all of the features and the random forest trained only on the top half of the features from the feature importance plot.